I know this is a 9900k thread but can someone please back up the (as far as I can see) unsubstantiated claims such as "AMD going to overtake Intel next year in speed, Mhz, IPC" "AMD going to crush Intel next year"... Etc.
I'm yet to see ANY evidence that says this is the case.
I don't doubt for a second that AMD will go mental next year and release a "16 core" consumer ryzen R7 or something mad along those lines. But I'm also willing to bet that Intel will STILL be king for gamers. What use is a 16core cpu to a gamer? AMD will keep slapping cores on and touting multithread render times in slides. But that's all they have at the moment. They can't beat Intel speed so it's... MOR CORES. I don't need em won't use em.
Most people on this forum have an interesting "I only care about gaming" perspective. You need to realise that that type of user isn't that common among the general desktop market.
Pretty much any benchmark will show their IPC as neck-and-neck currently, and Intel's IPC hasn't even moved since Skylake (which itself was only a 3% ish improvement on Broadwell). Obviously AMD has more cores at each price point, fewer security issues (that we know about), and the gaming picture looks a lot better now than it did 18 months ago when Ryzen was a brand new architecture. Even their energy efficiency is great, which is mainly because Intel are milking their main advantage: clock speeds.
Presentations slides from ages ago suggested that whilst Global Foundries' 14 nm process was designed for "low power" and optimised for 3 GHz speeds, the 7 nm process is designed for 5 GHz speeds. Whether that's possible whilst staying in a reasonable TDP envelope is yet to be seen but AMD seem confident. In terms of IPC, who knows. I've seen the 10-15% number floated around but no evidence for it.
Meanwhile Intel is stuck with tweaking their 14 nm process. Their maximum boost clocks have crept up over the last few Skylake refreshes whilst keeping the same TDP, and they seem to slightly bump their maximum overclocking potential with each generation too. AMD's core count has forced them to compete in that arena too, which is why the highest core count in their mainstream chips went from 4 to 6 last time around. It's now bumping again to 8, although they are also introducing a new tier which will have exclusivity of SMT, presumably to avoid cannibalising their HEDT line-up too much. So yeah, once again Intel will be introducing a new price bracket so they can charge more.
We know Intel won't have anything beyond "Coffee Lake 2" until 2H 2019 and let's be honest, that probably won't happen either (since 10 nm was originally due in 2016 and has been delayed half a dozen times already). I wouldn't be surprised if we see a Coffee Lake 3 next year, although what it'll look like is anyone's guess. They've been continuing with the ring bus design for their mainstream CPUs whilst they switched to mesh for their HEDT and Xeon line-up in 2015, but that was poorly received because overall performance was worse at the same clock speed. They're currently stretching the ring bus design to compete in the short-term but presumably they already know that isn't sustainable, otherwise they wouldn't have switched to using a mesh architecture for their higher core count chips. I believe the most cores ever seen in a single ring bus is 10, so something has to give. Surely eventually they'll need to embrace their mesh architecture and improve it enough that it at least matches AMD's infinity fabric design?
One thing is for sure: they cannot get away with their current strategy forever.